MyPowerPlant
Savings computed on-device
A Virtual Power Plant advisor for Victorian households: savings computed on-device, vendor recommendations, a chatbot and alerts.
Team project. I led the frontend and product flows — dashboard, onboarding, goals and the KPI maths — and worked on the FastAPI backend.
01 Problem
Victorian households deciding whether to join a Virtual Power Plant should see the savings maths, who to buy from, and when to act — without handing a retailer their meter data.
02 Approach
The savings model runs in the browser, so answers never leave the device. Around it sit a vendor recommendation path, a chatbot for the join-or-not questions and personalised alerts, with the UI built to WCAG AA. Server pieces run on FastAPI (wrapped with Mangum for Lambda) only where they are needed, backed by PostgreSQL with Alembic migrations. The stack is provisioned with Terraform — CloudFront over private S3, arm64 Lambda, ECR and SSM — and deployed through GitHub Actions OIDC with no long-lived AWS keys.
03 Result
A live advisory tool that computes savings on-device, recommends a vendor, answers follow-ups in the chatbot and can nudge a household with alerts, shipped across 100+ pull requests. There is no published conversion metric for it yet.
04 What didn’t work
Keeping the savings maths on-device rules out scoring every household input on the server; the chatbot and alerts work beside that constraint rather than replacing it.